Sei sulla pagina 1di 10

JOM

DOI: 10.1007/s11837-012-0350-0 2012 TMS

Soft Magnetic Materials in High-Frequency, High-Power Conversion Applications


ALEX M. LEARY,1,3 PAUL R. OHODNICKI,2 and MICHAEL E. MCHENRY1
1.Department of Materials Science and Engineering, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA. 2.Division of Chemistry and Surface Science, National Energy Technology Laboratory (NETL), 626 Cochrans Mill Road, Pittsburgh, PA 15236, USA. 3.e-mail: leary@cmu.edu

Advanced soft magnetic materials are needed to match high-power density and switching frequencies made possible by advances in wide band-gap semiconductors. Magnetics capable of operating at higher operating frequencies have the potential to greatly reduce the size of megawatt level power electronics. In this article, we examine the role of soft magnetic materials in high-frequency power applications and we discuss current materials limitations and highlight emerging trends in soft magnetic material design for high-frequency and power applications using the materials paradigm of synthesis structure property performance relationships.

INTRODUCTION Wilson1 denes power electronics as the technology associated with the efcient conversion, control, and conditioning of electric power by static means from its available input form into the desired electrical output form. Nearly all generated electricity requires some form of conversion. Power losses during conversion dissipate as heat during transmission and distribution to the end user and developments in soft magnetic materials resulted in signicant reductions in these losses over this time.2,3 Limited metering restricts precise measurement of transmission and distribution losses on a large scale.4 When generated commercial power does not exceed demand, the ratio of power sold to generated provides an estimate for these losses. Figure 1 shows the total electrical power generated for the retail market compared with the percentage that was not sold between 1949 to 2010 in the United States.5 In addition to a direct economic impact totaling $25.8 billion in lost revenue for a nominal retail price of $0.0988/kWh, the generation of this wasted power also results in additional greenhouse gas emissions that negatively impact the environment. These losses drive efciency improvements and are a lower bound of the total transmission, distribution, and conversion losses since they do not include losses experienced by the consumer downstream of the billing meter.

The Department of Energy expects demand for electricity to rise by 30% and estimates a $1.5 trillion cost to modernize the existing United States electricity infrastructure over the next 20 years.6 These upgrades and capacity expansions provide an opportunity to build and integrate new power electronics to meet demand while reducing waste and inefciencies. Movement toward grid-level integration of renewable energy sources and distributed storage systems requires new topologies to handle transient sources and facilitate two-way power conversion.7 Flexible alternating current (AC) Transmission Systems (FACTS) and High Voltage DC (HVDC) technologies aim to improve the efciency of power networks and benet from highfrequency conversion.8 Increases in DC power generation and loading also motivate research into new topologies containing high-frequency DCDC power converters.913 Independent of the power generation method, the laminated electrical steels traditionally used for power cores become inefcient at high switching frequencies. Additionally, soft magnetics used in power electronics can occupy signicant space, require extensive cooling, and limit designs. New large-scale systems must be cost competitive with existing systems and further cost reductions require materials advancements. These advancements will dictate the most successful topologies to take advantage of material strengths and minimize

Leary, Ohodnicki, and McHenry


4000

Electrical Power Generated (Billion kWh)

12 3500 3000 2500 2000 1500 1000 500 1950 1960 1970 1980 1990 2000 2010 6 Power Losses 8

10

Fig. 1. Total retail electrical power generated (NAICS 22) in the United States and estimated transmission and distribution losses, as measured by the ratio of power sold to power generated. Data taken from Ref. 5 includes independent power producers starting in 1989.

weaknesses. This article explores the use of soft magnetic materials in power conversion applications. The rst section describes the high-frequency limitations of soft magnetic materials, the Materials Survey section surveys the current state-of-theart materials, and the Future Processing Opportunities section highlights opportunities for future improvements. HIGH-FREQUENCY SWITCHING AND POWER CONVERSION Soft magnetic materials enable low loss inductive switching, which is useful in inductor, transformer and lter applications. The basic design challenge for inductive components becomes evident after considering Eq. 1, which relates Faradays law of induction to the voltage response of an ideal toroidal core with inductance L driven by an AC current, I = I0 sin(xt). V NA dB dI L LI0 x cosxt dt dt lN 2 A I0 x cosxt l

(1)

For constant maximum voltage (V), permeability (l), number of turns (N), and effective length (l), the cross-sectional area (A) is inversely proportional to the frequency (x = 2pf). This relation motivates the use of high-frequency switching to reduce size and weight of passive inductive components in power converters. However, nonlinear material properties limit scaling reductions in power magnetics, especially for applications above the kilowatt power range. Several studies examine these limitations to guide transformer and inductor design.1419 Additionally, high-frequency power converters rarely use purely sinusoidal currents to drive the magnetic

components. Instead, an active switching circuit modies the input signal using pulse width modulation by various techniques.20 Fourier analysis describes this modulated signal as the superposition of a many frequencies, so optimally designed soft magnetic materials must have broadband capability. The semiconductors used in active switching circuits have voltage and power limitations and highvoltage converters often divide the total output into manageable quantities with a multilevel circuit.21 Multilevel circuits can have the added benet of improving the harmonic quality of the drive current, but each additional level comes with a price as the semiconductors have associated switching and conduction losses. New SiC and GaN semiconductors have wider band gaps than currently available Si devices that result in much lower on-resistance.22 Lower losses in the active components of a power converter allow for higher power densities W=m3 : While several high-frequency converter designs operate without magnetics, considerations such as galvanic isolation and the scaling limitations associated with air core inductors suggest the need to further develop advanced magnetics.18 Currently, no commercial magnetic materials can match the performance level of the wide band gap semiconductors.23 A successful design minimizes the combined losses of the passive magnetic material, the windings, and the switching circuit for a given power output. Typically, loss mechanisms are complex and designers rely on empirical rather than analytical models. High-frequency losses in magnetic materials are dominated by classical and anomalous eddy currents caused by the motion of domain walls.24 Eddy currents generate heat by I2R losses in the magnetic material. For continuous operation, this heat must be dissipated through the component surface to prevent excessive temperatures. Scaling limitations result when surface area decreases while the amount of generated heat remains constant. Improvements in power density for a given power rating and efciency require increased cooling due to the reduced surface area available to smaller components. With currently available soft magnetic materials, thermal management limits converter power density below levels possible with advanced switching circuits. The steady-state temperature rise in the magnetic material for a given shape and power loss is a function of the materials thermal conductivity and emissivity, the local heat transfer conditions surrounding the material, and the surface area of the material exposed to these conditions. Scaling models describe the effects of size reductions for given design constraints and require descriptions of heat generation and thermal responses to losses. Passive components (inductors and capacitors) operate in concert with active components in a power converter and scaling relationships must consider both. A geometric factor, such

Losses (%)

Soft Magnetic Materials in High-Frequency, High-Power Conversion Applications

Fig. 2. Tape wound ring core geometry used for loss modeling.

as the area product25 or a component volume, combines the dimensions of passive and active components into a single variable. Here, we consider magnetic material property effects on scaling for the simple tape wound core geometry shown in Fig. 2 of outer and inner diameters (R1 and R2 ) and the core depth d. We choose the outer diameter of the ring core R1 as the geometric variable and x the inner diameter and depth to 0.7R1 and 0.3R1. The time averaged core loss is given by the wellknown variation of the Steinmetz Eq. 2 where k, a, and b are empirical ts to loss data. PC kf a Bb m (2)

Fig. 3. Power loss for R1 = 7 cm core for two relative permeabilities. Added power rating (orange) for a lower permeability material compared with additional winding losses (yellow) leads to higher efciency for constant core loss (Color gure online).

PW I2 RFr

where

Fr 1

2 2 6 p2 x2 l2 0 N n dc k (5) 2 768q b2 c

This loss, usually determined from a sinusoidal driving frequency, incorporates the hysteresis and eddy current losses due to changes of magnetization at a frequency f. The windings produce a eld H that is amplied by B = lH in the core. H /IN l (3)

where N is the number of turns over an effective length l and / is a geometric constant. We assume no DC bias and the useful portion of the induction is expressed as DB 2Bm . For a single layer of closely packed windings where l = Nw, winding losses for a given conductor resistivity (q) can be described as   1 Bm l 2 ql 2 PW Irms R (4) 2 l/N Ac The conductor cross section (Ac) decreases with increasing frequency due to the skin effect. Increasing N decreases the maximum drive current, but for high-frequency and power applications, smaller cores may limit the available winding space. Multiple winding layers lead to additional losses due to AC proximity effects. Sullivan26 accounts for these losses by dening Fr, a dimensionless factor that relates DC to AC resistance for a core wound with Litz wire.

The wire on the model core occupies a window with an areal dimension b2 c and uses Litz wire with n = 300 strands of diameter dc = 0.2 mm wire. We rst consider a case where Litz wire windings ll 2 the center core area with N pR2 2 =w and the wire crosssection is 50% copper q 2lX cm. Each wire occupies a square area w2 within the winding 2 window of size b2 c = p R2. To account for additional winding layers, the length per turn increases by 8w for each additional layer. The DC winding resistance is small compared with the AC resistance. A more detailed winding loss model for tape cores is found in Ref. 27. For magnetic materials with constant permeability over the ux range DB, the power rating (PVA) can be used to estimate the inductive efciency for a single core (gc) of volume Vc PVA Vrms Irms p/AlB2 dBH Vc B2 mf mf l dt 2l (6)

gc

PVA PC PW PVA

(7)

Assuming Steinmetz coefcients to be constant over the frequency range of interest, core loss is linear on a loglog loss versus frequency plot. Winding losses change slope with increased frequency due to skin and proximity effects. Figure 3 shows these losses and the output of a core with R1 = 7 cm and 308

Leary, Ohodnicki, and McHenry

Fig. 4. (a) Efciency versus frequency and (b) power density in kW=dm3 for the core geometry of Fig. 2.

turns for Bm = 0.5T with core loss values from Ref. 28. The two shaded areas show the difference between output power (VA) and total losses for two different relative permeabilities. We assume the same core loss for lower and higher permeability cores and discuss the assumption in the Materials Survey section. By lowering permeability, the material stores more inductive energy per cycle but also requires higher current (or more turns) to reach a desired induction. For this model, the additional stored inductive energy is greater than the added winding losses. If core loss is greater than winding loss, then materials with low permeability give efciency benets at high frequencies, as shown in Fig. 4a. For volume and weight limited applications, low permeability also gives high power density. Figure 4b shows the theoretical power density for this core geometry over a range of inductions and relative permeabilities. The 60 kW DCDC converter in Ref. 29 has a 40 kW=dm3 power density. Kolar et al.23 show a 10-fold increase in converter power density every 20 years. Published power density values vary depending on the inclusion of a cooling system. Magnetic components occupy a signicant portion of the volume. In the following, we extend this core model to estimate desirable magnetic material properties for high-frequency, high-power conversion. Reasonable designs must consider thermal limitations. In Refs. 14, 18, and 30, the relation between temperature rise and power loss relies on extrapolations from experimental data. We can also estimate the temperature rise DT for a material based on the thermal conductivity k as in Ref. 31. kD T PT h A (8)

The cooling system determines the heat transfer conditions at the surface, but the heat generated by the power loss PT = Pc + Pw must rst conduct

through the thermal path h within the material. We assume a cooling system to be able to remove PT from the surface at T1 and use the thermal parameter kDT to describe the maximum possible efciency of the inductive components in a 1 MVA power converter. Ring core geometries with many windings present a challenge because the windings limit heat transfer from the core. For this reason, we consider ring cores with three winding layers and dene the thermal path length as h = R1 R2. Each core and winding form a subcomponent within the converter and losses for the model subcomponents were calculated using the Monte Carlo method by randomly assigning model parameters over the following design space: (I) f 1 kHz ! 1 MHz; (II) Bm 0:5 ! 1:3T ; (III) lr 200 ! 5000; and (IV) R1 7 ! 20 cm. Core losses were calculated in mW=cm3 units from Eq. 2 using Finemet loss values from Ref. 28 with k = 3.935, a = 1.585, and b = 1.88 where the frequency is in kHz and the induction in Tesla. The measured core losses were accurate within 1% to 500 kHz. For converters that require multiple subcomponents to achieve a desired power rating, the connection arrangement effects the overall efciency (gt). Figure 5a shows the efciencies for individual subcomponents compared with the thermal parameter from Eq. 8. These efciency values correspond to the gt for subcomponents connected in parallel and successful designs have high efciencies and require low thermal parameters. The power ratings (PVA) form a banded pattern, and the 20 kVA and 80 kVA lines indicate the respective boundaries. For gc > 0.95, higher frequencies (Fig. 5b) and lower permeabilities (Fig. 5c) produced higher power ratings but require higher thermal parameters. The allowable temperature rise and thermal conductivity varies for different soft magnetic materials. Ferrites with k  4W=mK have operating limits below 200C

Soft Magnetic Materials in High-Frequency, High-Power Conversion Applications

Fig. 5. Performance measures for randomly designed sub-components with (a) efciency versus thermal parameter and power rating (PVA) versus thermal parameter. Higher frequencies (b) and lower permeabilities (c) yield higher PVA ratings.

1.00

MATERIALS SURVEY
2 Magnetic materials store inductive energy 1 2 LI and lter unwanted signals in power converters. The majority of heat generated in a circuit arises from this storage function and a good inductive material stores the most energy with the lowest loss. As shown in Eq. 1, L is largely a function of the permeability (l). Figure 7 shows the energy stored for different permeabilities. The area enclosed in the BH hysteresis curve for a material equals the core loss per cycle. The high induction levels approaching 2T commonly used at 50/60 Hz in Si steels are not achievable at higher frequency with currently available materials due to high coercivities (Hc). High-permeability materials (lr > 10,000) store very little energy prior to saturation. Materials with lower permeabilities store more energy per cycle at lower inductions but require higher driving elds and increased winding losses. Designers can achieve low-permeability cores by introducing an air gap or cutting a core made of highpermeability material. Cut cores introduce additional losses through ux fringing around the gaps that can also induce eddy currents in nearby conductive material.28 Poor manufacturing techniques greatly inuence cut core properties and can lead to much higher losses than expected.33 Materials with low permeability with respect to the driving eld direction are preferred in order to avoid complications due to cut cores. Cores with induced anisotropies transverse to the drive eld restrict domain wall motion and show low losses at high frequency.34 Alves et al.35 demonstrated improved performance in a low power yback converter design with a low permeability stress annealed Finemet core compared to a gapped ferrite core. The nanocomposite core was half the size of the ferrite core and produced 333% more power. For highpower, low-loss applications, low permeability is best achieved by inducing a controlled anisotropy perpendicular to the drive eld in a material that exhibits a high permeability and low coercivity as measured along the easy axis. Therefore, gures of merit for lowfrequency applications including high saturation

0.99

Efficiency

0.98

0.97

0.96
f > 40 kHz 20 kHz < f 40 kHz f 20 kHz

10

10

10

10

1 MVA Mass (kg)


Fig. 6. 1 MVA efciency versus total mass of soft magnetic material required. With the modeled material, higher frequencies reduce the amount of material required but result in lower efciencies.

and nanocomposites maintain low losses above 200C with k  9W=mK.32 The improved thermal properties of nanocomposite materials greatly expand the envelope of acceptable designs. The total converter power level and the individual core power rating determine the number of subcomponents required for a converter design. Figure 6 shows the converter efciency compared with the mass of soft magnetic material (density = 7.9 g/cm3) required to obtain 1 MVA. For the modeled material, higher frequencies reduce the required mass but sacrice efciency. This illustrates the need for improved high-frequency magnetics to maintain high efciency. A cost analysis is needed to determine the best overall design and is the subject of future work. The shape ratios used to dene R2 and d may not be optimal and were chosen based on assumed manufacturing limitations, such as ribbon width and structural integrity.

Leary, Ohodnicki, and McHenry

Fig. 7. Inductive energy storage for low-coercivity materials with different permeabilities. Low permeability materials store more energy per cycle but must maintain low coercivity to prevent losses.

magnetization and high permeability may not directly apply to power magnetic materials for relatively high frequency applications (10 kHz1 MHz), though high permeability prior to inducing anisotropy is a good predictor of low losses. Therefore, materials with tunable permeabilities through carefully designed thermomagnetic or thermomechanical processing allow designers to optimize the magnetic material to the active components. Several factors determine the performance of a magnetic material as a high-frequency inductor. The quality factor (Q) in Eq. 9 where l0 and l00 are the real and imaginary parts of complex permeability is a performance measure for an inductor. Q xL l0 00 R l (9)

The empirical constants are dened as in Ref. 28 to describe core loss in mW=cm3 . Materials with a high power ratio must store large amounts of power efciently. Although the actual amount of power stored in the material will differ due to the pulse width modulation and the duty cycle used, sinusoidal excitations are assumed here as most loss data is published using this waveform. Losses in soft magnetic materials consist of both static and dynamic hysteresis losses. To limit static hysteretic losses in bulk metal alloys such as silicon steels, metallurgists have sought to obtain large grain sizes in order to minimize domain wall pinning, lower coercivity, and hence reduce low-frequency losses. Hasegawa36 showed the advantages of magnetic materials with small crystalline grains within amorphous alloys. The random anisotropy model developed by Alben et al.37 for amorphous ferromagnets was applied by Herzer38 to nanocrystalline ferromagnets and showed that a reduction in grain size well below the ferromagnetic exchange length can lead to a rapid decrease in coercivity. The review of amorphous and nanocomposite soft magnets in Ref. 39 describes the properties of these alloys in more detail. Dynamic losses consist of classical eddy current and anomalous or excess losses. Bertotti40 describes the classical eddy current power loss in Eq. 11, where t is the material thickness and q is the resistivity, as Pcl
2 p2 t2 B2 mf 6q

(11)

High permeability materials saturate under low elds. For high power applications, the amount of energy stored must be considered as well as the losses, but the quality factor does not account for magnetic saturation. A gure of merit in Eq. 10 compares the inductive energy stored per unit volume during a half cycle to the measured losses for a sinusoidal driving eld during that period. Power Ratio
1 Stored Power 2 Bm H 2f  a Power Loss 1;000k f Bb 1;000 b B2 m

f 1;000

1a 10

kl0 lr

This quantity is valid for materials with constant permeability over the frequency range of interest.

This expression is valid when the material thickness is much less than the skin depth. Nanocomposite materials with q  130 lX cm and lr = 1,000 have a skin depth of 61 lm at 100 kHz and ribbons produced by rapid solidication are usually <30 lm thick suggesting that this expression is approximately valid. Engineering approaches to limiting these losses focus on reducing thickness with thin laminations and increasing resistivity. Figure 8 shows the power ratio from 10 for different soft magnetic materials described in Table I. Materials with poor performance at high frequencies, such as Si steels, and high permeability have low power ratios. The distributed air gap in powder cores allow for very low effective permeabilities and soft saturation, which can be benecial for some applications. Ferrites can also achieve low permeability, but their low saturation induction and thermal conductivity restricts their use in high-power converters. Nanocomposite alloys consist of nanocrystals precipitated in an amorphous matrix and offer an extended set of processing dimensions such as grain size and volume fraction to tailor properties. Nanocomposite alloys of the FINEMET type41 have been most intensively studied, and hence, they represent the majority of data points in Fig. 8. The highest power ratios in the nanocomposites exhibit low permeabilities

Soft Magnetic Materials in High-Frequency, High-Power Conversion Applications


100

9 8 7 6 5 4 3 2

Nanocomposite Ferrite Powder Amorphous Steel Cut Core Stress Annealed Field Annealed

Power Ratio

10

9 8 7 6 5 4 3 2

=2

00

r
1

=2

000
2 3 4 5

10

5 6 7 89

10

5 6 7 89

10

Frequency (Hz)
Fig. 8. Survey of published soft magnetic material performance. Dotted lines indicate Eq. 10 with k = 3.935, a = 1.585, and b = 1.888 for l = 200, 2000 at Bm = 0.1T.

thereby enhancing the maximum potential stored energy. In practice, this can be accomplished by cutting the core to induce an air gap or by engineering transverse anisotropy. Gapped Finemet cores42 and stress annealed material43,44 approach the calculated power ratio for the modeled material with low permeability and Bm = 0.1T as indicated. The eld annealed Co-rich FeCo nanocomposite in Ref. 45 shows comparable performance with stress annealed Fe-rich alloys.

Several published high frequency, high power converter designs demonstrate the advantages of nanocomposite core materials. Zhang et al.46 describe a 100-kW bidirectional DCDC power converter that switches at 25 kHz with 97.8% efciency. This design uses three Finemet cores that account for 32% of the total loss. The advantages of high induction can be seen in Ref. 17, where a 30-kW converter with a Finemet core showed three times higher power density than a ferrite design at

Table I. References and material descriptions from Fig. 8 Reference [12] [12] [12] [49] [49] [42] [42] [43] [47] [31] [44] [50] [50] [51] [51] [51] [45] [52] [53] 1 2 3 1 2 1 2 Author Li et al., 2010 Li et al., 2010 Li et al., 2010 Rylko et al., 2010 Rylko et al., 2010 Fukunaga et al., 1990 Fukunaga et al., 1990 Yanai et al., 2008 Long et al., 2008 Rylko et al., 2009 Fukunaga et al., 2002 Yoshida et al., 2000 Yoshida et al., 2000 Endo et al., 2000 Endo et al., 2000 Endo et al., 2000 Yoshizawa et al., 2003 Kolano-Burian et al., 2008 Martis and Rogers, 1994 Material MnZn ferrocube 3F51 NiZn ferrocube 4F1 Cool Mu METGLAS 2605SA1 JFE 10JNHF600 FINEMET FT-1M cut core Ferrite TDK H3ST cut core Stress annealed FINEMET Cut core FeCo HTX-002 MPP 26 powder Stress annealed FINEMET Amorphous Fe70 Al5 Ga2 P9:65 C5:75 B4:6 Si3 powder Mo-permalloy powder Mn-Zn ferrite Fe0:97 Cr0:03 76 Si0:5 B0:5 22 C2 amorphous powder Sendust Field annealed Fe8:8 Co70 Cu0:6 Nb2:6 Si9 B9 nanocomposite Field annealed Fe14:7 Co58:8 Cu1 Nb3 Si13:5 B9 nanocomposite METGLAS 2705M

1 2 1 2 3

Leary, Ohodnicki, and McHenry

200 kHz. A 25-kW converter with a FeCo nanocomposite core was 39% lighter than a Finemet-based design.47 Inoue shows a 6% efciency improvement for a Finemet core compared with a previous ferrite design and predicts improved efciencies with SiC switching.48 FUTURE PROCESSING OPPORTUNITIES Low resistivities of electrical steels and thermal conductivity of ferrites limit their high-frequency power applications. Of the various bulk materials, amorphous alloys and nanocrystalline/amorphous nanocomposites are best suited for high-frequency and high-power applications because of the combination of high resistivity, saturation induction, and thermal conductivity. Nanocomposites can be engineered with distinct advantages over amorphous alloys in terms of high-temperature stability making them superior candidates for long-term operation. The high Curie temperature of the amorphous phase in FeCo compositions54,55 allows for reduced cooling requirements and expands the allowable design space from Fig. 5. Conde et al.56,57 explore composition modications to FeCo nanocomposites to optimize microstructures. Processing considerations for nanocomposites are focused primarily on achieving thin and consistent ribbon cross sections through melt spinning or planar ow casting techniques as discussed in Ref. 58. Thin, high-quality ribbons allow for a reduction in eddy current losses and an improved packing factor that will enable designers to take advantage of the higher induction levels in these materials and reduce component size. Two additional areas where high-frequency losses and the power ratio of Eq. 10 can be improved for nanocomposite materials involve tuning permeability through advanced thermomechanical or thermomagnetic processing techniques and increasing resistivity through careful alloy selection. Because of the important role of the permeability in optimizing the performance of a magnetic core, a number of processing techniques has been devised to control the permeability for a given magnetic material. Adjustments to the annealing conditions produce hysteresis loops with constant permeability as a function of applied eld up to core saturation. The two typical methods are magnetic eld processing, where an external eld is applied in either a parallel or circumferential direction with respect to the core axis, and stress annealing in which the sample is placed under tension during crystallization. In both cases, a controlled anisotropy induces permeability changes while maintaining random anisotropy and low coercivity. Induced anisotropy transverse to the eld direction restricts domain wall movement and promotes low-loss rotational magnetization changes. For high-frequency and highpower applications, transverse induced anisotropy is desirable for minimizing anomalous eddy current losses and increasing the maximum stored energy.

Stress-induced anisotropy presents a promising method to tune anisotropy59 by magnetoelastic effects. However, while hardness is improved in Fe-based nanocomposite materials,60 they can become brittle after crystallization. Co-based compositions show improved mechanical properties61 and large responses to transverse eld annealing,62,63 but the cost of Co may preclude widespread use. It is important to point out, however, that highfrequency power converters require signicantly less material and may justify the use of more expensive compositions and processing techniques. The stress annealed materials in Fig. 8 illustrate the advantages of low permeability, but permeabilities <100 are not practical due to winding considerations. Research that focuses on mechanisms to limit eddy current losses presents the greatest opportunity to improve material performance. Resistivity and thermal conductivity are related to the atomic structure and in amorphous materials the disorder leads to increased electron scattering and increases resistivity. Composition adjustments to nanocomposite materials can further effect resistivity by impacting grain size.64 Very high resistivities are correlated with decreased thermal conductivities that limit ferrite materials from use in high-power applications. This thermal conductivity limitation may also apply to nanogranular-based material approaches where an insulating oxide phase is continuous.19 The kinetics of nanocrystallization and the sequence of crystallization events impact the properties of nanocomposite materials.65 Recent work has demonstrated that large isotropic pressures on the order of 1 GPa can have a pronounced impact on the kinetics of crystallization from amorphous precursors. In some cases, a large applied pressure enhances crystallization kinetics,66,67,68 while in other cases, it tends to inhibit crystallization.69,70 Crystallization at elevated pressures is relevant for compaction to form bulk parts of arbitrary shape from rapidly solidied ribbons of amorphous metals. As a result, most fundamental investigations of the effects of large isotropic pressure on crystallization have focused on Al-based compositions with attractive mechanical properties. In Ref. 68, the effects of applied pressure on crystallization were described by using the specic nucleation work (W) shown in Eq. 12, which is the sum of the activation energy for the formation of a critical cluster (DG ) and the activation energy barrier for diffusion (Qn). In the classical nucleation theory, the interface energy of a critical cluster is calculated as the product of an interfacial energy (r) and the total area of the critical nucleus while the volume energy of formation is comprised of the thermodynamic driving force DG, the strain energy (E), and the product of the pressure and the molar volume change between the amorphous and crystalline phases. The crystalline phase has a lower volume than the amorphous phase. Qn is related to the atomic mobility in Ref. 71

Soft Magnetic Materials in High-Frequency, High-Power Conversion Applications

by the Arrhenius relation M Mo expQn =kT and shows pressure dependance.72 W DG Qn 16pr3
a V c DG E2 3PVm m

Qn (12)

Materials able to operate at high temperatures with a high thermal conductivity expand the allowable design conditions by reducing cooling requirements and tolerating high power losses. ACKNOWLEDGMENTS The work of A.M. Leary and M.E. McHenry was supported by ARPA-E Award Number DE-AR0000219 and the ARL through Grant No. W911NF-08-20024. Disclaimer: This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specic commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reect those of the United States Government or any agency thereof. REFERENCES
1. T.G. Wilson, IEEE Trans. Power Eletron. 15, 439 (2000). 2. I.S. Jacobs, J. Appl. Phys. 50, 7294 (1979). 3. L. Johnson, IEEE Trans. Power Apparatus Syst. 1, 68 (1982). 4. C.A. Dortolina and R. Nadira, IEEE Trans. Power Syst. 20, 1119 (2005). 5. Department of Energy, Electricity Overview, 19492010. http://www.eia.gov/totalenergy. 6. Department of Energy, What the Smart Grid Means to Americas Future. http://www.smartgrid.gov. 7. W. Kramer, S. Chakraborty, B. Kroposki, and H. Thomas, Advanced Power Electronic Interfaces for Distributed Energy Systems Part 1: Systems and Topologies. Technical Report March (Golden, CO: National Renewable Energy Laboratory, 2008). 8. G.F. Reed, B.M. Grainger, H. Bassi, E. Taylor, Z.-H. Mao, and A.K. Jones (Paper presented at the Panel Session on FACTS Applications To Improve Power System Dynamic Performance, IEEE PES T&D Conference and Exposition, April 2010), pp. 110. 9. S. Kouro, M. Malinowski, K. Gopakumar, J. Pou, L.G. rez, and J.I. Leon, Franquelo, B. Wu, J. Rodriguez, M.A. Pe IEEE Trans. Indust. Electron. 57, 2553 (2010). 10. A.R. Hefner Jr. (Paper presented at Power and Energy Society General MeetingConversion and Delivery of Electrical Energy in the 21st Century, August 2008 IEEE, 2008), pp. 12. 11. H. Fan and H. Li, Proc. Twenty-Fifth Annual IEEE Applied Power Electronics Conference and Exposition (APEC) (Piscatay, NJ: IEEE, February 2010), pp. 210215. 12. Q. Li, M. Lim, J. Sun, A. Ball, Y. Ying, F.C Lee, and K.D.T. Ngo, Proc. Twenty-Fifth Annual IEEE Applied Power Electronics Conference and Exposition (APEC) (Piscataway, NJ: IEEE, May 2009), pp. 533539. 13. W. van der Merwe and T. Mouton, Proc. 2009 IEEE International Conference on Industrial Technology (Piscataway, NJ: IEEE, 2009), pp. 16.

Pressure may be a useful processing variable to effect phase selection in nanocomposite materials by shifting the cluster formation energy of different phases. Co-rich Fe,Co-based compositions contain several crystalline phases with comparable free energies after crystallization in contrast to the Fe-rich compositions in which only body-centered cubic Fe-rich nanocrystalline phases are observed.73,74,75 Pressure may also be useful to rene grain size by decreasing the nucleation barrier. At higher pressures, restrictions on atomic mobility may limit grain growth allow for compositions with less early transition metals that are typically added to restrict diffusion.76 In addition to isotropic pressures, large anisotropic stresses can be exploited at elevated temperatures by traditional metallurgical hot rolling. No reports have been identied in which hot rolling techniques have been applied to crystallization of melt spun amorphous ribbons for soft magnetic applications. Because large anisotropic stresses can be generated during hot rolling, a potential opportunity exists for the development of novel anisotropic microstructures and stress-induced magnetic anisotropy in a practical and technologically relevant manner. In the case of Co-rich alloys, the anisotropy in the compliance matrices of both cubic and hexagonal phases may lead to preferential crystallization by aligning the soft axis to the directions of high stress. High densities of planar defects such as stacking faults have also been found in the nanocrystalline phases of these alloys which may provide a unique source of induced magnetic anisotropy during such anisotropic thermomechanical processing techniques.77 Stress annealing amorphous ribbons is a well-known phenomenon, but we are unable to nd published studies concerning rolling techniques for soft magnetic nanocomposite ribbons. The effects of hot and cold rolling present a potential research opportunity to determine the potential benets of pressure on the soft magnetic properties. CONCLUSION The adoption of high-frequency power converters for MW scale application requires advances in soft magnetic materials. The design optimization here points to the need for low loss materials with low, linear permeabilities that remain constant at high frequencies. FeCo nanocomposites are good candidates for these applications due to their high Curie temperatures, response to stress and eld annealing, and good thermal conductivity of tape wound cores.

Leary, Ohodnicki, and McHenry


14. W.G. Odendaal and J.A. Ferreira, IEEE Trans. Indust. Appl. 35, 932 (1999). 15. C. Xiao (Ph.D. dissertation, Virginia Polytechnic Institute and State, Blacksburg, VA, 2006). 16. J.W. Kolar, U. Drofenik, J. Biela, M.L. Heldwein, H. Ertl, T. Friedli, and S.D. Round, Proc. Power Conversion ConferenceNagoya (Piscataway, NJ: IEEE, April 2007), pp. 929. 17. W. Shen, F. Wang, D. Boroyevich, and W. Tipton, IEEE Trans. Indust. Appl. 44, 213 (2008). 18. D.J. Perreault, J. Hu, J.M. Rivas, Y. Han, O. Leitermann, R.C.N. Pilawa-Podgurski, A. Sagneri, and C.R. Sullivan, Proc. 2009 Twenty-Fourth Annual IEEE Applied Power Electronics Conference and Exposition (Piscataway, NJ: IEEE, 2009), pp. 114. 19. C.R. Sullivan, Proc. 2009 IEEE Custom Integrated Circuits Conference (Piscataway, NJ: IEEE, 2009), pp. 291298. 20. G. Walker and G. Ledwich. IEEE Trans. Power Electron. 14, 74 (1999). 21. M. Malinowski, K. Gopakumar, J. Rodriguez, and M.A. rez, IEEE Trans. Indust. Electron. 57, 2197 (2010). Pe 22. J.L. Hudgins, G.S. Simin, E. Santi, and M.A. Khan, IEEE Trans. Power Electron. 18, 907 (2003). 23. J.W. Kolar, J. Biela, S. Wafer, T. Friedli, and U. Badstuebner, Proc. 6th International Conference on Integrated Power Electronics Systems (CIPS) (Piscataway, NJ: IEEE, 2010), pp. 1618. 24. K. Suzuki, A. Makino, A. Inoue, and T. Masumoto, J. Appl. Phys. 74, 331 (1993). 25. W.T. McLyman, Transformer and Inductor Design Handbook, 2nd ed. (New York: Marcel Dekker, Inc., 1988). 26. C.R. Sullivan, IEEE Trans. Power Electron. 14, 283 (1999). 27. G. Lefevre, H. Chazal, J.P. Ferrieux, and J. Roudet, Application of Dowell method for Nanocrystalline toroid high frequency transformers (Paper presented at 35th Annual IEEE Power Electronics Specialists Conference, number 1, June 2004), pp. 899904. 28. W. Shen, F. Wang, D. Boroyevich, and C.W. Tipton, IEEE Trans. Power Electron. 23, 475 (2008). 29. M Pavlovsky, Y. Tsuruta, and A. Kawamura, Proc. IEEE Energy Conversion Congress and Exposition (Piscataway, NJ: IEEE, September 2009), pp. 17681774. 30. W.-J. Gu and R. Liu (Paper presented at 4th Annual IEEE Power Electronics Specialists Conference, Seattle, WA, June 1993). 31. M.S. Rylko, K.J. Hartnett, J.G. Hayes, and M.G. Egan, Proc. IEEE Applied Power Electronics Conference and Exposition (Piscataway, NJ: IEEE, February 2009), pp. 20432049. 32. Y. Wang, S.W.H. De Haan, and J.A. Ferreira, Proc. EPE, i (2009). 33. T.E. Salem, D.P. Urciuloli, V. Lubormirsky, and G.K. Ovrebo, Proc. Twenty Second Annual IEEE Applied Power Electronics Conference (Piscataway, NJ: IEEE, February 2007), pp. 12581263. 34. J.D. Livingston and W.G. Morris, J. Appl. Phys. 57, 3555 (1985). 35. F. Costa, F. Alves, J.B. Desmoulins, D. Herisson, and J.F. Rialland, Proc. IEEE 31st Annual Power Electronics Specialists Conference (June 2000), pp. 308313. 36. R. Hasegawa, J. Optoelectron. Adv. Mater. 6, 503 (2004). 37. R. Alben, J.J. Becker, and M.C. Chi, J. Appl. Phys. 49, 1653 (1978). 38. G. Herzer, IEEE Trans. Magn. 26, 1397 (1990). 39. M.E. McHenry, M.A. Willard, and D.E. Laughlin, Progr. Mater. Sci. 44, 291 (1999). 40. G. Bertotti, IEEE Trans. Magn. 24 (1988). 41. Y. Yoshizawa, S. Oguma, and K. Yamauchi, J. Appl. Phys. 64, 6044 (1988). 42. H. Fukunaga, T. Eguchi, K. Koga, Y. Ohta, and H. Kakehashi, IEEE Trans. Magn. 26, 2008 (1990). 43. T. Yanai, K. Takagi, K. Takahashi, M. Nakano, Y. Yoshizawa, and H. Fukunaga, J. Magn. Magn. Mater. 320, 833 (2008). 44. H. Fukunaga, T. Yanai, H. Tanaka, M. Nakano, K. Takahashi, Y. Yoshizawa, K. Ishiyama, and K.I. Arai, IEEE Trans. Magn. 38, 3138 (2002). 45. Y. Yoshizawa, S. Fujii, D.H. Ping, M. Ohnuma, and K. Hono, Scr. Mater. 48, 863 (2003). 46. J. Zhang, J.-S. Lai, R.-Y. Kim, and W. Yu, IEEE Trans. Power Electron. 22, 1145 (2007). 47. J. Long, M. McHenry, D.P. Urciuoli, V. Keylin, J. Huth, and T.E. Salem, J. Appl. Phys. 103, 07E705 (2008). 48. S. Inoue and H. Akagi, IEEE Trans. Power Electron. 22, 535 (2007). 49. M.S. Rylko, J.G. Hayes, and M.G. Egan, Proc. 2010 IEEE Vehicle Power and Propulsion Conference (Piscataway, NJ: IEEE, 2010), pp. 17. 50. S. Yoshida, T. Mizushima, T. Hatanai, and A. Inoue, IEEE Trans. Magn. 36, 3424 (2000). 51. I. Endo, H. Tatsumi, I. Otsuka, H. Yamamoto, A. Shintani, H. Koshimoto, M. Yagi, and K. Murata, IEEE Trans. Magn. 36, 3421 (2000). 52. A. Kolano-Burian, R. Kolano, J. Szynowski, and L.K. Varga, J. Magn. Magn. Mater. 320, e758 (2008). 53. R.J. Martis and D.W. Rogers, Proc. Applied 791 Power Electronics Conference and Exposition, APEC 94 (Piscataway, NJ: IEEE, February 1994), pp. 233237. 54. P.R. Ohodnicki, S.Y. Park, D.E. Laughlin, M.E. McHenry, V. Keylin, and M.A. Willard, J. Appl. Phys. 103, 07E729 (2008). 55. M.E. McHenry, F. Johnson, H. Okumura, T. Ohkubo, V.R.V. Ramanan, and D.E. Laughlin, Scr. Mater. 48, 881 (2003). 56. C.F. Conde, A. Conde, P. Svec, and P. Ochin, Mater. Sci. Eng. A 375377, 718 (2004). vec, zquez, A. Conde, P. S 57. C.F. Conde, J.M. Borrego, J.S. Bla kovic , J. Alloy. Compd. 509, 1994 (2011). D. Janic 58. S. Shen, P. Ohodnicki, S. Kernion, A. Leary, V. Keylin, J. Huth, and M.E. McHenry, TMS, 2012. 59. M. Ohnuma, G. Herzer, P. Kozikowski, C. Polak, V. Budinsky, and S. Koppoju, Acta Mater. 60, 1278 (2012). 60. C.-Y. Um, F. Johnson, M. Simone, J. Barrow, and M.E. McHenry, J. Appl. Phys. 97, 10F504 (2005). 61. T.M. Heil, K.J. Wahl, A.C. Lewis, J.D. Mattison, and M.A. Willard, Appl. Phys. Lett. 90, 212508 (2007). 62. P.R. Ohodnicki, J. Long, D.E. Laughlin, M.E. McHenry, V. Keylin, and J. Huth, J. Appl. Phys. 104, 113909 (2008). 63. P.R. Ohodnicki, V. Keylin, H.K. McWilliams, D.E. Laughlin, and M.E.McHenry, J. Appl. Phys. 103, 07E740 (2008). 64. M.S. Lucas, W.C. Bourne, A.O. Sheets, L. Brunke, M.D. Alexander, J.M. Shank, E. Michel, S.L. Semiatin, J. Horwath, and Z. Turgut, Mater. Sci. Eng. B 176, 1079 (2011). 65. A. Hsiao, M.E. McHenry, D.E. Laughlin, M.J. Kramer, C. Ashe, and T. Ohkubo, IEEE Trans. Magn. 38, 3039 (2002). 66. F. Ye and K. Lu, Phys. Rev. B 60, 7018 (1999). 67. B. Varga, A. Lovas, F. Ye, X.J. Gu, and K. Lu, Mater. Sci. Eng. A 286, 193 (2000). 68. Y.X. Zhuang, J.Z. Jiang, T.J. Zhou, H. Rasmussen, and L. Gerward, Appl. Phys. Lett. 77, 4133 (2000). 69. F. Ye and K. Lu, Acta Metall. 46, 5965 (1998). 70. X.J. Gu, H.J. Jin, H.W. Zhang, J.Q. Wang, and K. Lu, Acta Mater. 45, 1091 (2001). 71. J.-O. Andersson and J. Agren, J. Appl. Phys. 72, 1350 (1992). 72. F. Faupel, W. Frank, M.-P. Macht, V. Naundorf, K. Ratzke, H. Schober, S. Sharma, and H. Teichler, Rev. Modern Phys. 75, 237 (2003). 73. M.A. Willard, T.M. Heil, and R. Goswami, Metall. Mater. Trans. A 38A, 725 (2007). 74. P.R. Ohodnicki Jr., D.E. Laughlin, M.E. McHenry, and M. Widom, Acta Mater. 58, 4804 (2010). 75. S.J. Kernion, P.R. Ohodnicki, and M.E. McHenry, J. Appl. Phys. 111, 07A316 (2012). 76. S.J. Kernion, K.J. Miller, S. Shen, V. Keylin, J. Huth, and M.E. McHenry, IEEE. Trans. Magn. 47, 3452 (2011). 77. P.R. Ohodnicki, Y.L. Qin, M.E. McHenry, D.E. Laughlin, and V. Keylin, J. Magn. Magn. Mater. 322, 315 (2010).

Potrebbero piacerti anche